r2e-edits/deepswe-verifier-qwen25-14b-merged-v1

DeepSWE Verifier - Qwen2.5 Coder 14B Merged

This model is a merged version of LoRA adapters trained for verification tasks with the Qwen2.5-Coder-14B-Instruct base model.

Model Details

• Base Model: Qwen/Qwen2.5-Coder-14B-Instruct
• LoRA Adapters: deepswe-verifier-only-matching-pairs-v1
• Model Type: Merged model (LoRA adapters integrated into base model)
• Use Case: Software engineering verification tasks

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "r2e-edits/deepswe-verifier-qwen25-14b-merged-v1",
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
)

tokenizer = AutoTokenizer.from_pretrained(
    "r2e-edits/deepswe-verifier-qwen25-14b-merged-v1",
    trust_remote_code=True
)

Training Details

This model was created by merging LoRA adapters trained on verification tasks with the base Qwen2.5-Coder-14B-Instruct model.